Quantitative evaluation and modeling of twodimensional neovascular network complexity: the surface fractal dimension
 Fabio Grizzi^{1, 2}Email author,
 Carlo Russo^{1, 2},
 Piergiuseppe Colombo^{3},
 Barbara Franceschini^{1, 2},
 Eldo E Frezza^{4},
 Everardo Cobos^{5} and
 Maurizio ChirivaInternati^{6}
DOI: 10.1186/14712407514
© Grizzi et al; licensee BioMed Central Ltd. 2005
Received: 15 October 2004
Accepted: 08 February 2005
Published: 08 February 2005
Abstract
Background
Modeling the complex development and growth of tumor angiogenesis using mathematics and biological data is a burgeoning area of cancer research. Architectural complexity is the main feature of every anatomical system, including organs, tissues, cells and subcellular entities. The vascular system is a complex network whose geometrical characteristics cannot be properly defined using the principles of Euclidean geometry, which is only capable of interpreting regular and smooth objects that are almost impossible to find in Nature. However, fractal geometry is a more powerful means of quantifying the spatial complexity of real objects.
Methods
This paper introduces the surface fractal dimension (D _{ s } ) as a numerical index of the twodimensional (2D) geometrical complexity of tumor vascular networks, and their behavior during computersimulated changes in vessel density and distribution.
Results
We show that D _{ s }significantly depends on the number of vessels and their pattern of distribution. This demonstrates that the quantitative evaluation of the 2D geometrical complexity of tumor vascular systems can be useful not only to measure its complex architecture, but also to model its development and growth.
Conclusions
Studying the fractal properties of neovascularity induces reflections upon the real significance of the complex form of branched anatomical structures, in an attempt to define more appropriate methods of describing them quantitatively. This knowledge can be used to predict the aggressiveness of malignant tumors and design compounds that can halt the process of angiogenesis and influence tumor growth.
Background
The term "angiogenesis" defines the fundamental process of the development and growth of new blood vessels from the preexisting vasculature, and is essential for reproduction, development and wound repair [1]. Under these conditions, it is highly regulated: i.e. "turned on" for brief periods of time (days) and then completely inhibited.
The cyclic nature of the microvascular bed in the corpus luteum provides a unique experimental model for examining the discrete physiological steps of angiogenesis in the life cycle of endothelial cells which, together with pericytes (supportive vascular smooth muscle cells), carry all of the genetic information necessary to form tubes, branches and entire capillary networks.
However, many human diseases (including solid tumors) are driven by persistently upregulated angiogenesis [1]. In some nonmalignant processes, such as pyogenic granuloma or keloid formation [2], angiogenesis is prolonged but still selflimited; however, this is not true of tumor angiogenesis which, once begun, continues indefinitely until the entire tumor is eradicated or the host dies. Without blood vessels, tumors cannot grow beyond a critical size (1–2 mm) or metastasize to another organ.
Angiogenesis is one of the most complex dynamic processes in biology, and is highly regulated by a balance of pro and antiangiogenic molecules. It is now widely accepted that the "angiogenic switch" is "off" when the effects of proangiogenic molecules is balanced by that of antiangiogenic molecules, and "on" when the net balance is tipped in favor of angiogenesis [1, 3]. Pro and antiangiogenic molecules can be secreted from cancer cells, endothelial cells, stromal cells, blood, and the extracellular matrix [4, 5], the relative contributions of which are likely to change with tumor type and site, as well as with tumor growth, regression and relapse [1].
Although considerable advances have been made in our molecular and cellular knowledge of the promotion, mediation and inhibition of angiogenesis, very little is known about its underlying complex dynamics. Vasculature and more generally tubular organs develop in a wide variety of ways involving many cell processes [6–8].
In mathematical terms, angiogenesis is a nonlinear dynamic system that is discontinuous in space and time, but advances through qualitatively different states. The word state defines the configuration pattern of the system at any given moment, and a dynamic system can be represented as a set of different states and a number of transitions from one state to another over a certain time interval [9, 10].
Tumor vessels are structurally and functionally abnormal [1, 3]: unlike normal vessels, they are highly disorganized, tortuous and dilated, and have uneven diameters, and excessive branching and shunts. This may be mainly due to the heterogeneous distribution of angiogenic regulators, such as vascularendothelial growth factor (VEGF), basic fibroblastic growth factor (bFGF) and angiopoietin [5, 13], leading to chaotic tumor blood flow, and hypoxic and acidic tumoral regions [5, 14–16]. Moreover, although it is commonly believed that the endothelial cells makingup tumor vessels are genetically stable, diploid cells (and thus different from genetically unstable neoplastic cells), tumor vasculature seems to be much more unpredictable [17].
These conditions all reduce the effectiveness of treatments, modulate the production of pro and antiangiogenic molecules, and select a subset of more aggressive cancer cells with higher metastatic potential [1].
A large number of clinical trials of antiangiogenic therapies are being conducted throughout the world, but investigators are still concerned about how to achieve the maximum benefit from them and how to monitor patient response. There are currently no markers of the net angiogenic activity of a tumor that can help investigators to design specific antiangiogenic treatment strategies [5, 18], but it is reasonable to resume that the quantification of various aspects of tumor vasculature may provide an indication of angiogenic activity.
One oftenquantified element of tumor vasculature is microvessel density (MVD), which is used to allow a histological assessment of tumor angiogenesis. The results of studies carried out over the last decade have suggested the value of using tumor MVD as a prognostic index in a wide variety of solid cancers, and it has also recently been assumed that MVD may reveal the degree of angiogenic activity in a tumor. On the basis of these assumptions, the quantification of MVD is thought to be a surrogate marker of the efficacy of antiangiogenic agents as well as a means of assessing which patients are good candidates for antiangiogenic therapy. However, MVD has a number of substantial limitations, mainly due to the complex biology characterizing tumor vasculature [17], and the highly irregular geometry that the vascular system assumes in real space, which cannot be measured using the principles of Euclidean geometry because it is only capable of interpreting regular and smooth objects that are almost impossible to find in Nature.
However, quantitative descriptors of its geometrical complexity can be usefully abstracted from the fractal geometry introduced by Benoit Mandelbrot in 1975 [20, 21]. We here discuss the surface fractal dimension (D _{ S } ) as a quantitative index of the 2D geometrical complexity of vascular networks and their behavior during computersimulated changes in vessel density and distribution.
Geometrical properties of a vascular network
The human vascular system can be geometrically depicted as a complex fractal network of vessels that irregularly branch with a systematic reduction in their length and diameter [19].
Fractal objects are mainly characterized by four properties: a) the irregularity of their shape; b) the selfsimilarity of their structure; c) their noninteger or fractal dimension; and d) scaling, which means that the measured properties depends on the scale at which they are measured [22].
One particular feature of fractal objects is that the schemas defining them are continuously repeated at decreasing orders of magnitude, and so the form of their component parts is similar to that of the whole [20, 21]: this property is called selfsimilarity. Unlike geometrical selfsimilarity, which only concerns mathematical fractal objects in which every smaller piece is an exact duplicate of the whole (e.g. Koch's snowflake curve, Sierpinski's triangle and Menger's sponge), statistical selfsimilarity concerns all complex anatomical systems, including tumor vasculature. The smaller pieces constituting anatomical entities are rarely identical copies of the whole, but more frequently "similar" to it and, in such systems, the statistical properties of the pieces are proportional to the statistical properties of the whole [23].
Benoit Mandelbrot uses the symbol D _{ γ }to indicates the topological dimension, and the symbol D to indicate that of HausdorffBesicovitch (also called the fractal dimension). The D _{ γ }and D of all Euclidean figures are coincident (D _{ γ } = D), but this is not true of fractal objects in which D is always >D _{ γ }.
Anatomical structures are also hierarchical systems that operate at different spatial and temporal scales, and different patterns can change, appear or disappear depending on the scale of magnification [22]. A fundamental characteristic is that the process operating at a given scale cannot be important at higher or lower scales [23].
The irregularity and selfsimilarity underlying scale changes are the main attributes of the architectural complexity of both normal and pathological biological entities [22–26]. In other words, the shape of a selfsimilar object does not change when the scale of measure changes because every part of it is similar to the original object; however, the magnitude and other geometrical parameters (e.g. the outline perimeter) of an irregular object differ when inspected at increasing resolutions that reveal an increasing number of details [25]. Over the last decade, accumulating experimental evidence has shown that the fractal patterns or selfsimilar structures of biological tissues can only be observed within the scaling window of an experimentally established measure of length ε _{1}ε _{2} (Fig. 4), within which experimental data sets follow a straight line with a slope (1D): i.e. the fractal dimension remains invariant at different magnifications [20–27].
Methods
Computeraided modeling of twodimensional vascular tree complexity
We have developed a computer model to simulate the geometrical complexity of a histological twodimensional section of a tumor vascular tree that automatically generates an unlimited number of images with a changeable density of vessels irregularly distributed on a planar surface.
One thousand images were automatically generated for each vessel density (from five to 50 vessels, with the number being increased by five in each group), and their D _{ S }were estimated using the boxcounting method [22].
D _{ S }was automatically estimated using the equation:
where ε is the sidelength of the box, and N(ε) the smallest number of boxes of side ε required to completely contain the irregular object (Fig. 4).
As the zero limit cannot be applied to biological images, D _{ S }was estimated by means of the equation:
(2) D _{ S }= d
where d is the slope of the graph of log [N(ε)] against log (1/ε), in a fixed range of sidelengths (ε _{1}ε _{2}) empirically evaluated by visualization [20–29].
Statistical analysis
All of the data are expressed as mean values ± standard deviation, and the results were analysed using the Statistica software package (StatSoft Inc. Tulsa, USA). Unvaried analysis was performed by means of the Student t as required for parametric variables. p values of less than 0.05 were considered statistically significant.
Results
Discussion and conclusions
One of the most important and distinctive characteristics of biological systems is the complexity of their shape (geometrical or spatial complexity) and functions (behavioral complexity). Complexity is a real quality of organized biological matter that is mainly manifested in the living world as diversity and organization. No two anatomical systems are exactly alike because of the enormous variability not only between the different members of a population, but also between the component parts of an organism. The word complexity has long been used descriptively in order to describe, for example, a large number of genes or cellular interconnections [33], but complexity can also reside in the structure of a system (i.e. an intricate architecture or the existence of many different component parts with varying interactions) or its nonlinear functions (i.e. physiological rhythms are rarely strictly periodic but fluctuate irregularly over time) [34].
The vascular system is a complex network consisting of branched tubes of different sizes that are irregularly settled in the surrounding environment [6, 7]. This geometrical characteristic highlights the complexity of its generating process in space and time, and greatly biases any quantitative method that tends to idealize it as a smooth and regular Euclidean object.
However, both normal and tumor vasculature can more properly be considered fractal objects because of their irregular shape (spatial conformation), selfsimilar structure, noninteger dimension and dependence on the scale of observation (scaling effect) [19, 35–37].
We here discuss the estimate of D _{ S }as a quantitative index of the 2D spatial complexity of the vascular tree, in order to provide a closertoreality measure of this complex anatomical entity (Figs. 3 and 4).
The theory underlying D _{ S }was abstracted from fractal geometry, which is also called the geometry of irregularity [20, 21]. The concept of spatial conformation has played a fundamental role in the study of biological macromolecules in chemistry (particularly biochemistry) since the early 1950s. However, it has only been introduced in the science of morphology as theoretical morphology, which studies extant organismal forms (complex structures of interdependent and subordinate elements whose relationships and properties are largely determined by their function in the whole) as a subset of the range of theoretically possible morphologies [32].
The significance of D _{ S }also comes from the fact that, like any other complex biological system, the vascular tree cannot be correctly quantified by measuring its individual properties (i.e. microvessel density, MVD). D _{ S }is a parameter that depends on: a) the number of vessels; b) the spatial relationships between the vascular components; and c) the interactions between the vascular components and the surrounding environment. In other words, its estimate is "ecologically" important because it provides a quantitative index of the "habitat structure".
As computer models are crucial for scientific procedures, and the modeling process itself represents the hypotheticaldeductive approach in science [30–32], we developed a simple computeraided model capable of generating an unlimited number of 2D images of a simulated vascular network. The model was simplified by using a minimum amount of mathematical complexity and only two variables: the number of vessels and their pattern of distribution. A total of 10,000 images showing a different number of unconnected vessels irregularly distributed on a planar surface were automatically generated (Fig. 5) and, interestingly, it was found that D _{ S }increased with the number of vessels making up the system (Fig. 6); furthermore, its value changed when the same number of vessels were differently distributed in the surrounding environment.
In other words, it is plausible that an equal number of vessels may have different spacefilling properties depending on their distribution pattern. These results suggest the usefulness of this model when comparing real vasculature configurations in order to explore the morphological variability that can be produced in nature, as it is now well known that aberrant vascular architectures in tumors may affect the uniform delivery of specific drugs to all cancer cells [15].
The model also suggests that:
a) D _{ s }can be an estimate of the 2D geometrical complexity of the vascular system. As 2D vascular complexity depends on the number of vessels and their distribution pattern, the use of MVD quantification alone to measure the angiogenic dependence of a tumor is strongly biased because the number of vessels does not reflect the number of tumor cells that can be supported by a vessel. Moreover the metabolic needs of cancer cells vary with the tissue of origin and change with tumor progression [18].
b) D _{ S }depends on the degree of vessel contiguity and continuity. These two geometrical properties determine what is called the intercapillary distance, and are not only involved in the spatial complexity of tumor vasculature, but also reflect the inviolable demand of a growing tumor for sufficient levels of nutrition and oxygen exchange. Intercapillary distances are locally defined by the net balance between pro and antiangiogenic molecules in each microtissue region, as well as by nonangiogenic factors such as the oxygen and nutrient consumption rates of tumor cells. In normal tissue, vessel density fairly accurately reflects cell metabolic demands because evolutionary pressures have led to close and efficient coupling between vascular supply and metabolic needs. In tumors, the close coupling between vascular density and oxygen or nutrient consumption (i.e. the environment) may be loosened [18], thus altering not only the number of vessels but also the whole vascular architecture [15, 38].
c) D _{ S }falls between 0 (corresponding to the Euclidean dimension of a point) and 2 (the dimension of a plane). The more D _{ S }tends towards 2, the more the analyzed vascular configuration tends to fill a 2D space and the greater its geometrical complexity.
In conclusion, the present study indicates that the complex geometry of tumor vasculature and its wellknown biological characteristics [18] mean that vascular network cannot be measured on the basis of MVD estimates alone. These findings also support the findings of various authors who have shown the uselessness of MVD as a predictor of antiangiogenic treatment efficacy or for stratifying patients in therapeutic trials [14, 39–41].
Scientific knowledge develops through the evolution of new concepts, and this process is usually driven by new methodologies that provide previously unavailable observation. The potential broad applicability of the proposed quantitative index makes it possible to explore the range of the morphological variability of vasculature that can be produced in nature, thus increasing its diagnostic importance in cancer research.
Abbreviations
 Ds:

Surface fractal dimension
 2D:

twodimensional
 VEGF:

Vascularendothelial growth factor
 bFGF:

basic fibroblastic growth factor
 MVD:

microvessel density.
Declarations
Authors’ Affiliations
References
 Carmeliet P: Angiogenesis in health and disease. Nature Medicine. 2003, 9: 653670. 10.1038/nm0603653.View ArticlePubMedGoogle Scholar
 Woolf N: Pathology. Basic and Systemic. 1998, WB Saunders Company, LondonGoogle Scholar
 Fidler IJ, Ellis LM: Neoplastic angiogenesis – not all blood vessels are created equal. New England Journal of Medicine. 2004, 351: 215216. 10.1056/NEJMp048080.View ArticlePubMedGoogle Scholar
 Fukumura D, Xavier R, Sugiura T, Chen Y, Park EC, Lu N, Selig M, Nielsen G, Taksir T, Jain RK, Seed B: Tumor induction of VEGF promoter activity in stromal cells. Cell. 1998, 94: 715725. 10.1016/S00928674(00)817316.View ArticlePubMedGoogle Scholar
 Tandle A, Blazer DG, Libutti SK: Antiangiogenic gene therapy of cancer: recent developments. Journal of Translational Medicine. 2004, 2: 2210.1186/14795876222.View ArticlePubMedPubMed CentralGoogle Scholar
 Hogan BL, Kolodziej PA: Organogenesis: molecular mechanisms of tubulogenesis. Nat Rev Genet. 2002, 3: 51323. 10.1038/nrg840.View ArticlePubMedGoogle Scholar
 Lubarsky B, Krasnow MA: Tube morphogenesis: making and shaping biological tubes. Cell. 2003, 112: 1928. 10.1016/S00928674(02)012837.View ArticlePubMedGoogle Scholar
 Parker LH, Schmidt M, Jin SW, Gray AM, Beis D, Pham T, Frantz G, Palmieri S, Hillan K, Stainier DY, De Sauvage FJ, Ye W: The endothelialcellderived secreted factor Egfl7 regulates vascular tube formation. Nature. 2004, 428: 754758. 10.1038/nature02416.View ArticlePubMedGoogle Scholar
 Abraham R: Complex dynamics. 1991, Aerial Press, Santa CruzGoogle Scholar
 Abraham R, Shaw C: Dynamics, The geometry of behavior. 1992, AddisonWesley, Reading MA, SecondGoogle Scholar
 Folkman J, Klagsbrun M: Angiogenic factors. Science. 1987, 235: 4427.View ArticlePubMedGoogle Scholar
 ColvilleNash PR, Willoughby DA: Growth factors in angiogenesis: current interest and therapeutic potential. Molecular Medicine Today. 1997, 3: 1423. 10.1016/S13574310(96)100484.View ArticlePubMedGoogle Scholar
 Jain RK, Schlenger K, Hockel M, Yuan F: Quantitative angiogenesis assays: progress and problems. Nature Medicine. 1997, 3: 12031208. 10.1038/nm11971203.View ArticlePubMedGoogle Scholar
 Harris AL: Hypoxia – a key regulatory factor in tumor growth. Nat Rev Cancer. 2002, 2: 3847. 10.1038/nrc704.View ArticlePubMedGoogle Scholar
 Munn LL: Aberrant vascular architecture in tumors and its importance in drugbased therapies. Drug Discovery Today. 2003, 8: 396403. 10.1016/S13596446(03)026862.View ArticlePubMedGoogle Scholar
 Stergiopulos N, Porret CA, De Brouwer S, Meister JJ: Arterial vasomotion: effect of flow and evidence of nonlinear dynamics. Am J Physiol. 1998, 274: H185864.PubMedGoogle Scholar
 Streubel B, Chott A, Huber D, Exner M, Jager U, Wagner O, Schwarzinger I: Lymphomaspecific genetic aberrations in microvascular endothelial cells in Bcell lymphomas. New England Journal of Medicine. 2004, 351: 250259. 10.1056/NEJMoa033153.View ArticlePubMedGoogle Scholar
 Hlatky L, Hahnfeldt P, Folkman J: Clinical application of antiangiogenic therapy: microvessel density, what it does and doesn't tell us. Journal National Cancer Institute. 2002, 94: 883893.View ArticleGoogle Scholar
 Grizzi F, Colombo P, Barbieri B, Franceschini B, Roncalli M, ChirivaInternati M, Muzzio PC, Dioguardi N: Correspondence re: E. Sabo et al., Microscopic analysis and significance of vascular architectural complexity in renal cell carcinoma. Clinical Cancer Research. 2001, 7: 33053307.PubMedGoogle Scholar
 Mandelbrot BB: Les objets fractals: forme, hasard et dimension. 1975, Flammarion, ParisGoogle Scholar
 Mandelbrot BB: The Fractal Geometry of Nature. 1982, Freeman, San FranciscoGoogle Scholar
 Grizzi F, Franceschini B, ChirivaInternati M, Hermonat PL, Shah G, Muzzio PC, Dioguardi N: The complexity and the Microscopy in the anatomical sciences. Science, Technology and Education of Microscopy: an Overview. 2003, Formatex, SpainGoogle Scholar
 Bassingthwaighte JB, Liebovitch LS, West BJ: Fractal physiology. 1994, Oxford University Press, New YorkView ArticleGoogle Scholar
 Losa GA: Fractal morphometry of cell complexity. Rivista di Biologia. 2002, Biology Forum, 95: 239250.Google Scholar
 Nonnenmacher TF, Baumann G, Barth A, Losa GA: Digital image analysis of selfsimilar cell profiles. Int J Biomed Comput. 1994, 37: 131138.View ArticlePubMedGoogle Scholar
 Losa GA, Nonnenmacher TF: Selfsimilarity and fractal irregularity in pathologic tissues. Mod Pathol. 1996, 9: 174182.PubMedGoogle Scholar
 Dollinger JW, Metzler R, Nonnemacher TF: Biasymptotic fractals: Fractals between lower and upper bounds. J Phys A: Math Gen. 1998, 31: 38393847. 10.1088/03054470/31/16/012.View ArticleGoogle Scholar
 Paumgartner D, Losa G, Weibel ER: Resolution effect on the stereological estimation of surface and volume and ist interpretation in terms of fractal dimension. Journal of Microscopy. 1981, 121: 51View ArticlePubMedGoogle Scholar
 Cross SS: Fractals in pathology. J Pathol. 1988, 182: 18. 10.1002/(SICI)10969896(199705)182:1<1::AIDPATH808>3.3.CO;22.View ArticleGoogle Scholar
 Massoud TF, Hademenos GJ, Young WL, Gao E, PileSpellman J, Vinuela F: Principles and philosophy of modeling in biomedical research. FASEB Journal. 1998, 12: 275285.PubMedGoogle Scholar
 Weiss JN, Qu Z, Garfinkel A: Understanding biological complexity: lessons from the past. FASEB Journal. 2003, 17: 16. 10.1096/fj.020408rev.View ArticlePubMedGoogle Scholar
 McGhee GR: Theoretical morphology: the concept and its applications. 1998, Columbia University Press, New YorkGoogle Scholar
 Szathmary E, Jordan F, Csaba P: Can genes explain biological complexity?. Science. 2001, 292: 13151316. 10.1126/science.1060852.View ArticlePubMedGoogle Scholar
 Golbeter A: Biochemical oscillations and cellular rhythms. The molecular bases of periodic and chaotic behavior. 1996, Cambridge University PressView ArticleGoogle Scholar
 Baish JW, Jain RK: Cancer, angiogenesis and fractals. Nature Medicine. 1998, 4: 98410.1038/1952.View ArticlePubMedGoogle Scholar
 Baish JW, Jain RK: Fractals and cancer. Cancer Research. 2000, 60: 36833688.PubMedGoogle Scholar
 Bassingthwaighte JB, King RB, Roger SA: Fractal nature of regional myocardial blood flow heterogeneity. Circulation Research. 1989, 65: 578590.View ArticlePubMedPubMed CentralGoogle Scholar
 Padera TP, Stoll BR, Tooredman JB, Capen D, di Tomaso E, Jain RK: Pathology: cancer cells compress intratumor vessels. Nature. 2004, 427: 69510.1038/427695a.View ArticlePubMedGoogle Scholar
 Uzzan B, Nicolas P, Cucherat M, Perret GY: Microvessel density as a prognostic factor in women with breast cancer: a systematic review of the literature and metaanalysis. Cancer Research. 2004, 64: 29412955.View ArticlePubMedGoogle Scholar
 Hansen S, Sorensen FB, Vach W, Grabau DA, Bak M, Rose C: Microvessel density compared with the Chalkley count in a prognostic study of angiogenesis in breast cancer patients. Histopathology. 2004, 44: 428436. 10.1111/j.13652559.2004.01848.x.View ArticlePubMedGoogle Scholar
 Fox SB: Tumor angiogenesis and prognosis. Histopathology. 1997, 30: 294301. 10.1046/j.13652559.1997.d01606.x.View ArticlePubMedGoogle Scholar
 The prepublication history for this paper can be accessed here:http://www.biomedcentral.com/14712407/5/14/prepub
Prepublication history
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.